I conducted a survey to test several DV. I assumed that several questions in the survey (e.g. liking of product, repeated purchase) load on one factor/DV (e.g. willingness to buy).
To test whether these two questions really ‘load’ on the same construct, a principle component analysis (PCA = form of exploratory factor analysis) is carried out via IBM SPSS. I got significant results here and extracted several factors.
My question now: Is it necessary to carry out a Confirmatory factor analysis as well? Do I need to test the model I received (containing the specific amount of factors)?
I would be grateful for your feedback,
Anna
Dear Anna-Maria,
Like Ertuğrul I agree that Exploratory Factor Analysis (EFA) is more relevant than Confirmatory Factor Analysis when you have designed your own instrument (which I assume you have done).
I think the issue is whether you are trying to establish several scales or only one scale. If it is the former then I agree with Ertuğrul that it would be wise to start with an EFA. If it is the latter then you can start with a Principal Component Analysis in order to validate your single scale. This is possible even when your sample size is quite small.
Have a look at my two study guides:
https://www.researchgate.net/publication/304490328_Advice_on_Exploratory_Factor_Analysis
https://www.researchgate.net/publication/280936182_Advice_on_Reliability_Analysis_with_Small_Samples
Technical Report Advice on Reliability Analysis with Small Samples
Working Paper Advice on Exploratory Factor Analysis
Dear Anna-Maria,
yes of course
i am attached the google book link please seen and download it.
Applied multivariate statistics for the social sciences: Analyses with SAS and IBM's SPSS(KA Pituch, JP Stevens - 2016 - books.google.com)
Best wishes to your future research
Thank you for your feedback! This is really helpful..
What I still don't get:
- What is the difference between VARIMAX, Promax or direct oblimin rotation?
- How can I perform a Confirmatory Factor analysis in SPSS? Is it the calculation of Construct reliablity, AVE etc.?
Thank you!
Anna
Dear Anna,
When you constructed your questioner, I assume that you had had a clear relation between the items in your questioner and the constructs to be measured; i.e. which items measure which constructs. In this case, I suggest you to directly evaluate your questioner by CFA. It will help you to avoid indeterminacy problem in EFA. In addition, you may have a more elaborate measurement/factor model in CFA. Basically, an EFA model is a special model of CFA. EFA is very restrictive.
You may refer to chapter 6 and 7 of Bollen, K.A. (1989) "Structural Equation with Latent Variables".
Moreover, before you apply a factor model (EFA or CFA), you should consider the relationships between the items and the constructs; whether they are reflective or formatif. Factor models are only applicable to reflective relations.
Best,
Yusep
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS